Google Professional Cloud Security Engineer Question 361
Select 4Google Cloud PlatformYour organization has deployed a machine learning model on Google Cloud AI Platform. The model processes sensitive customer data to provide predictions. To secure the AI workload, you are tasked with implementing measures to protect the data during processing and while the model is being served. Which of the following actions should you take to meet these security requirements?
- A
Enable VPC Service Controls to restrict access to the AI Platform and its APIs.
- B
Use Customer-Managed Encryption Keys (CMEK) to encrypt data at rest used by the AI model.
- C
Enable the AI Platform model to log all predictions and input data for auditing purposes.
- D
Implement IAM permissions to ensure only authorized users and service accounts can access the AI model.
- E
Use Confidential Computing to encrypt data in use during model inference.
Show answer and explanation
Correct answers: A, B, D, E
Explanation
Securing AI workloads requires a multi-faceted approach to protect data in transit, at rest, and in use. VPC Service Controls restrict resource access, CMEK protects data at rest, IAM ensures proper access control, and Confidential Computing secures data during processing. Logging predictions, while useful for auditing, does not directly contribute to securing the AI workload and can expose sensitive information if mishandled.
- A. Correct.
Enabling VPC Service Controls enhances security by restricting access to AI Platform resources and APIs, reducing the risk of data exfiltration.
- B. Correct.
Using CMEK ensures that sensitive data at rest is encrypted with keys controlled by your organization, providing an additional layer of security.
- C. Incorrect.
Logging predictions and input data can expose sensitive information if not handled securely. While auditing is important, this option is not about securing the AI workload directly.
- D. Correct.
Implementing IAM permissions ensures that only authorized users and service accounts can access the AI model, protecting it from unauthorized access.
- E. Correct.
Confidential Computing encrypts data in use, ensuring that data remains protected even during model inference, which is critical when dealing with sensitive information.